Ollama vision models: - `llama3.2-vision` (https://github.com/ollama/ollama/releases/tag/v0.4.0-rc5) - `llava` and variants